Optimal freight rate management for VLCC Jonas Røn Thorsgaard Steen S070573 Kongens Lyngby 2013 IMM-M.Sc.-2013-13 Technical University of Denmark Informatics and Mathematical Modelling Building 321, DK-2800 Kongens Lyngby, Denmark Phone +45 45253351, Fax +45 45882673 [email protected] www.imm.dtu.dk IMM-M.Sc.-2013-13 Summary (English) The objective of the thesis is to develop a stochastic framework as a practical decision support tool for managing VLCC chartering, and to analyse the effi- ciency of such a framework. The efficient managing process, determined by the modelling framework, is within the usage of FFA contracts in fixing future prices on voyage contracts. Prices on voyage contracts are determined by a volatile spot market, which can be hedged using FFA’s. The work is divided into three different parts with the main purpose of developing a stochastic programming model. The three parts are divided into a presentation of the freight rate market and financial derivatives, the development of a statistical framework making pre- dictions in spot rates and the development of a stochastic programming model making allocation decisions. The scope of the thesis is limited by the lack of previous prepared studies in the dirty tanker freight rate market. When it comes to the usage of financial derivatives in the tanker market, studies are almost none existent. Only very few studies has been made to introduce the derivative market. The thesis therefore introduces the derivative market in the form of introducing Forward Freight Agreements and FFA Options. It is examined how the financial derivatives are structured, with the purpose of disseminating knowledge of financial derivatives in the market, and to examine the efficiency and limitations in using derivatives, such as lack of liquidity. Fur- thermore, the entire set-up of doing VLCC chartering is examined to give the reader the understanding of the chartering process. The work is based on the Dirty Tanker 3 (TD3) index published on the Baltic Exchange, in the form of Worldscale points. A minor statistical analysis is made on the dirty tanker index with the purpose ii of developing a statistical framework in which to make reasonable predictions on the future level of freight rates. There has not yet been made any studies that determines a precise way to make predictions on tanker freight rates. It has therefore been chosen to make a well known time series analysis on the TD3 index. It has been examined how autoregressive processes and ARMA-GARCH processes performs in the freight rate market. It seems that none of these processes perform significantly better than a simple bootstrap method. The bootstrap method has therefore been chosen in the thesis as the most adequate choice of making predictions on the TD3 index, even though it is a very naive way of doing predictions. The final work of the thesis is to develop a stochastic programming framework, in which to make optimal decisions on how to manage VLCC chartering. The framework adopted in the thesis is a well known decision model in financial engineering, proposed by Stavros as a CVaR programming model. The model optimizes the expected income of the chosen strategy, and minimizes the down side risk exposure in terms of Conditional Value at Risk. The results from implementing and analysing the framework on prior historical records seem rea- sonable. It is very clear that introducing FFA in the managing process is indeed controlling risk exposures in a positive way. All strategies including FFA’s seem to outperform strategies without using FFA’s, both in risk exposure and in expected income. Summary (Danish) Målet for dette speciale er at udarbejde en stokastisk metode til på en praktisk og effektiv måde at supportere administration af VLCC udlejning. Desuden er formålet at analysere resultater fra den udviklede metode for at vurdere gra- den af forbedringer, et rederi kan opnå i indtjening af VLCC udlejning. Den optimale administrationsproces, bestemt af den udarbejdede metode, er inden for brugen af FFA kontrakter til at fixe fremtidige priser på voyage kontrakter. Priser på voyage kontrakter er bestemt af et volatilt spotmarked, som kan blive hedged med brugen af FFA. Udarbejdelsen er opdelt i 3 dele, med det hoved- formål at udvikle en stokastisk programmeringsmodel. Specialet er begrænset af den betydeligt manglende udarbejdelse af forudgående studier inden for frag- tratemarkedet for dirty tankers. Når det kommer til brugen af financielle derivater i tankermarkedet, er forud- gående studier nærmest ikke eksisterende. Det har kun været muligt at finde 2 tidligere studier, der introducere derivatmarkedet for fragt rater. Specialet intro- ducerer derfor derivatmarkedet i form af FFA og FFA optioner. Det gennemgås, hvordan derivater er opbygget, med hendblik på at udbrede kendskabet til fi- nancielle derivater i tankermarkedet. Desuden gennemgås effekten fra brugen af derivater, samt begrænsningerne i markedet, såsom manglende likviditet. Der- udover gennemgås fragt rate markedet, for at introducere læseren til, hvordan udlejning af VLCC foregår. Specialet er baseret på Dirty Tanker 3 (TD3) index, som offentliggøres på Baltix Exchange, i form af Worldscale point. Herefter foretages en mindre statistisk analyse på tankerindexet, med henblik på at udvikle en statistisk metode til at lave fornuftige forudsigelser omkring den fremtidige udvikling af fragtratemarkedet. Der har ikke tidligere været udarbej- iv det studier, der på en effektiv og præcis måde har kunne forudsige fragtraterne. Det er derfor blevet valgt at lave en velkendt tidsserie analyse på TD3 indexet. Det er blevet analyseret hvordan autoregresive processer og ARMA-GARCH processer præsterer i fragtratemarkedet. Det er vurderet, at ingen af tidsseri- eprocesserne analyseret i denne afhandling præsterer beytdeligt bedre end et simpelt bootstrap. Det er derfor valgt at bruge et bootstrap sample til at lave forudsigelser på TD3 indexet selvom det er velkendt at, bootstrapmetoden er en naiv måde at lave forudsigelser. Den afsluttende del af specialet omhandler udarbejdelsen af en stokastisk pro- grammeringsmodel, hvorfra den optimale beslutningsproces omkring admini- stration af udlejning af VLCC er bestemt. Modellen, som er brugt i specialet er en velkendt beslutningsmodel inden for financial ingeniørvidenskab, foreslået af Stavros som en CVaR programmerings model. Modellen optimerer over den forventede indtjening af den valgte strategi, og minimerer risikoeksponeringen i form af Conditional Value at Risk. Resultaterne fra at implementere og ana- lysere den stokastiske metode på historiske data virker yderst rimelige. Det ses tydeligt, at brugen af FFA kontrakter ved styring af VLCC udlejning giver en positiv mulighed for at kontrollere risikoeksponeringen. Alle undersøgte strate- gier med FFA’er klarer sig betydeligt bedre end strategier, hvor FFA kontrakter er udeladt, både når det angår risikoeksponering men også den forventede ind- komst. Preface This thesis was prepared at the department of Informatics and Mathematical Modelling at the Technical University of Denmark in fulfilment of the require- ments for acquiring an M.Sc. in Informatics. The thesis deals with optimal decision management in the dirty tanker business, of managing freight rates on very large crude carriers. The thesis consists of an introduction into the market of freight rates and finan- cial derivatives. Furthermore, it consist of a statistical analysis of the freight rates, with the purpose of developing predictions. Finally it consists of the de- velopment and analysis of a stochastic programming model, supporting the need for optimal decision making within the dirty tanker chartering. Lyngby, 01-March-2013 Jonas Røn Thorsgaard Steen vi Acknowledgements I would like to thank my two supervisors Kourosh Marjani Rasmussen and Lasse Engbo Christiansen for helpful supervisions and constructive feedback, when needed. Furthermore I would like to thank Kenneth Juhls and Carsten Dreyer Christensen from the risk management department at Maersk Oil Trading, for being helpful in sharing business knowledge, in defining the framework of the thesis and to provide data materials. Finally I would like to thank Christina Steen for making here time to read and make suggestions for corrections in the entire thesis. Without the help from all these people the thesis would not have got in to where it is today. viii Notation Sets, indices and general notation used throughout the thesis. T = ft0; t1; : : : ; tτ ; : : : ; tT g set of time periods, from today t0 until maturity tT . Unless stated otherwise in the text all time periods are of equal duration which is typically taken to be one month. Ω = f1; 2;:::;Ng index set of scenarios. Σt = f1; 2;:::;Stg index set of states in economy at period t. Γl index set of random numbers i for scenario l. K={spot,ffa_m0,. ,ffa_q5} set of contracts that can be used by the shipping company t index of time periods from the set T . l index of scenario from the set Ω. i index of observation in historical data set. κ index of contracts from the set K. T number indicating planning horizon. N number of scenarios. λ 2 (0; 1) number determine the risk profile in the stochastic programming model. x α number determine the confidence interval of the downside risk. pi observed price in the historical data set at index i. ri rate of price development from index i − 1 to index i. di date at index i. Xl scenario l. t Xl state of scenario l at time t. pl probability of scenario l. tτ t al item in scenario state Xl , indicating the average of spot prices in the period from tτ−1 to tτ . t t bl item in scenario state Xl , indicating the price of settling a spot contract at time t. tτ t fl item in scenario state Xl , indicating the price of a FFA contract purchased at time t0 and settled on the average spot prices in the period from tτ−1 to tτ tτ µκ,l income of contracts κ at time tτ .